diff --git a/docs/installation/configure-cdi.md b/docs/installation/configure-cdi.md
new file mode 100644
index 00000000..65d79ba3
--- /dev/null
+++ b/docs/installation/configure-cdi.md
@@ -0,0 +1,146 @@
+---
+title: Enable NVIDIA CDI support for HAMi
+sidebar_label: NVIDIA CDI support
+---
+
+This page explains what CDI does and which Helm chart values enable HAMi's NVIDIA CDI integration.
+
+## What is CDI?
+
+[CDI (Container Device Interface)](https://github.com/cncf-tags/container-device-interface) is a specification for describing the OCI configuration that a container needs to use a device. A CDI specification can define device nodes, mounts, environment variables, and OCI hooks in JSON or YAML.
+
+CDI is vendor-neutral. HAMi currently supports CDI device injection only for NVIDIA GPUs. The HAMi configuration and troubleshooting guidance on this page do not apply to devices from other vendors.
+
+CDI identifies each device by a fully qualified name:
+
+```text
+vendor.com/class=device-name
+```
+
+When a container runtime receives this name, it finds the matching CDI specification in a directory such as `/etc/cdi` or `/var/run/cdi` and applies the specified changes to the container's OCI runtime specification.
+
+CDI does not schedule devices, allocate resources, or enforce quotas. In HAMi's NVIDIA CDI integration, the scheduler and NVIDIA Device Plugin continue to select and allocate GPUs; CDI injects the allocated GPUs and their runtime requirements into containers.
+
+## What problems does CDI solve?
+
+Injecting a complex device often requires more than mounting a single `/dev` node. A GPU container may also need driver libraries, several device nodes, environment variables, and lifecycle hooks. Without a common interface, vendors must adapt this logic to individual container runtimes, and runtimes may need vendor-specific implementations.
+
+CDI moves these OCI changes into a common specification. It provides:
+
+- A standard format for device nodes, mounts, environment variables, and hooks.
+- Stable device names for passing allocation results from Device Plugins to container runtimes.
+- Less dependency on runtime-specific device injection implementations.
+- An inspectable record of the device configuration applied to a container.
+
+## How HAMi uses CDI for NVIDIA GPUs
+
+With `cdi-annotations` enabled for the NVIDIA Device Plugin, HAMi uses the following injection flow:
+
+1. The HAMi scheduler and NVIDIA Device Plugin select and allocate a GPU.
+2. The Device Plugin generates a CDI specification and writes it to `/var/run/cdi` on the node.
+3. The Device Plugin returns the allocated device name through a CDI annotation.
+4. A CDI-aware container runtime resolves the annotation and applies the matching CDI specification to the container's OCI runtime specification.
+
+HAMi uses the following CDI kind for NVIDIA GPUs:
+
+```text
+k8s.device-plugin.nvidia.com/gpu
+```
+
+## Prerequisites
+
+Before enabling NVIDIA CDI support in HAMi, confirm that:
+
+- The NVIDIA driver and NVIDIA Container Toolkit are installed on each GPU node.
+- The container runtime supports CDI and has CDI enabled.
+- The container runtime can read CDI specifications from `/var/run/cdi`.
+- The actual NVIDIA driver root and `nvidia-ctk` path on each node are known.
+
+For runtime-specific instructions, see the [upstream CDI configuration guide](https://github.com/cncf-tags/container-device-interface#how-to-configure-cdi).
+
+## Configure the Helm chart
+
+Save the following values as `values-cdi.yaml`:
+
+```yaml
+devicePlugin:
+  deviceListStrategy: "cdi-annotations"
+  nvidiaDriverRoot: "<driver-root>"
+  nvidiaHookPath: "<nvidia-ctk-path>"
+```
+
+- `devicePlugin.deviceListStrategy`: Set this to `cdi-annotations` so that the HAMi Device Plugin passes device names through CDI annotations.
+- `devicePlugin.nvidiaDriverRoot`: The NVIDIA driver root on the node. HAMi uses this directory to discover driver files and device nodes.
+- `devicePlugin.nvidiaHookPath`: The `nvidia-ctk` executable used on the node. HAMi writes this path into the generated CDI specification.
+
+Replace the two path placeholders according to the deployment:
+
+- Host-installed NVIDIA driver and Container Toolkit: use `/` for `<driver-root>` and `/usr/bin/nvidia-ctk` for `<nvidia-ctk-path>`.
+- GPU Operator-managed driver and Container Toolkit: use `/run/nvidia/driver` for `<driver-root>` and `/usr/local/nvidia/toolkit/nvidia-ctk` for `<nvidia-ctk-path>`.
+
+:::warning
+
+The two path values must match the actual node filesystem. Do not copy values from a different deployment model. `nvidiaHookPath` must point to an executable that is available to the container runtime on the node.
+
+:::
+
+Install or upgrade HAMi:
+
+```bash
+helm upgrade --install hami hami-charts/hami \
+  --namespace kube-system \
+  --create-namespace \
+  --values values-cdi.yaml
+```
+
+## Verify the configuration
+
+First, confirm that Helm applied the expected values:
+
+```bash
+helm get values hami -n kube-system
+```
+
+On a GPU node, check the selected `nvidia-ctk` executable and the CDI specification generated by HAMi. The following example uses the host-installed path; replace `NVIDIA_CTK_PATH` in GPU Operator environments:
+
+```bash
+NVIDIA_CTK_PATH=/usr/bin/nvidia-ctk
+sudo test -x "$NVIDIA_CTK_PATH"
+"$NVIDIA_CTK_PATH" --version
+sudo ls -l /var/run/cdi
+sudo jq . /var/run/cdi/k8s.device-plugin.nvidia.com-gpu.json
+```
+
+Confirm that:
+
+- The CDI kind is `k8s.device-plugin.nvidia.com/gpu`.
+- The allocated GPU UUID is present in the device list.
+- Every device node and hook path in the specification exists on the node.
+
+Finally, check the Device Plugin logs:
+
+```bash
+kubectl logs -n kube-system \
+  -l app.kubernetes.io/component=hami-device-plugin \
+  --tail=200 | grep -i cdi
+```
+
+The logs should not contain errors about CDI specification generation, unresolved devices, or missing hook paths.
+
+## Troubleshooting
+
+### `failed to stat CDI host device "/dev/nvidia-modeset": no such file or directory`
+
+`devicePlugin.nvidiaDriverRoot` does not match the driver layout on the node. Correct the value using the path rules above, upgrade HAMi, and inspect the regenerated CDI specification.
+
+### `stderr: No help topic for 'disable-device-node-modification'`
+
+The NVIDIA Container Toolkit executable referenced by the CDI specification does not support the `disable-device-node-modification` command. This command requires NVIDIA Container Toolkit 1.18.0 or later.
+
+Upgrade NVIDIA Container Toolkit and verify both the hook path recorded in the CDI specification and the version of the executable at that path. Host-installed and GPU Operator-managed Toolkit installations are independent; upgrading one does not update the other.
+
+[GPU Operator v25.10.1](https://github.com/NVIDIA/gpu-operator/blob/v25.10.1/deployments/gpu-operator/values.yaml#L220-L224) uses NVIDIA Container Toolkit v1.18.1. For related details, see [Toolkit version parser broken for casual kind demo user](https://github.com/kubernetes-sigs/dra-driver-nvidia-gpu/issues/568).
+
+### `error running createContainer hook #0: fork/exec /usr/bin/nvidia-ctk: no such file or directory`
+
+The hook path recorded in the CDI specification does not exist on the node. Locate the actual `nvidia-ctk` executable, correct `devicePlugin.nvidiaHookPath`, and upgrade HAMi to regenerate the CDI specification.
diff --git a/docs/installation/how-to-use-hami-dra.md b/docs/installation/how-to-use-hami-dra.md
index 2695115e..6db129c5 100644
--- a/docs/installation/how-to-use-hami-dra.md
+++ b/docs/installation/how-to-use-hami-dra.md
@@ -18,7 +18,7 @@ HAMi has provided support for K8s [DRA](https://kubernetes.io/docs/concepts/sche
 ## Prerequisites
 
 - Kubernetes version >= 1.34 with DRA Consumable Capacity [featuregate](https://kubernetes.io/docs/reference/command-line-tools-reference/feature-gates/) enabled
-- [CDI](https://github.com/cncf-tags/container-device-interface?tab=readme-ov-file#how-to-configure-cdi) must be enabled in the underlying container runtime (such as containerd or CRI-O)
+- [CDI](./configure-cdi.md) must be enabled in the underlying container runtime (such as containerd or CRI-O)
 - NVIDIA GPU Driver 440 or later
 
 ## Installation
diff --git a/docs/installation/prerequisites.md b/docs/installation/prerequisites.md
index 22d26669..59a4e880 100644
--- a/docs/installation/prerequisites.md
+++ b/docs/installation/prerequisites.md
@@ -17,6 +17,8 @@ This guide assumes pre-installation of NVIDIA drivers and the `nvidia-container-
 
 For details see [Installing the NVIDIA Container Toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html).
 
+To inject NVIDIA GPUs through CDI, follow [Enable NVIDIA CDI support for HAMi](./configure-cdi.md) before installing HAMi. Verify the container runtime, driver root, and NVIDIA Container Toolkit path first.
+
 ### Example for debian-based systems with Docker and containerd
 
 #### Install the `nvidia-container-toolkit`
diff --git a/i18n/zh/docusaurus-plugin-content-docs/current/installation/configure-cdi.md b/i18n/zh/docusaurus-plugin-content-docs/current/installation/configure-cdi.md
new file mode 100644
index 00000000..6ccdae28
--- /dev/null
+++ b/i18n/zh/docusaurus-plugin-content-docs/current/installation/configure-cdi.md
@@ -0,0 +1,147 @@
+---
+title: 为 HAMi 启用 NVIDIA CDI 支持
+sidebar_label: NVIDIA CDI 支持
+translated: true
+---
+
+本文介绍 CDI 的作用，以及为 HAMi 配置 NVIDIA CDI 支持所需的 Helm Chart 参数。
+
+## CDI 是什么
+
+[CDI（Container Device Interface）](https://github.com/cncf-tags/container-device-interface)是一项容器设备接口规范。它使用 JSON 或 YAML 文件描述容器使用某个设备时需要应用的 OCI 配置，包括设备节点、文件挂载、环境变量和 OCI hook。
+
+CDI 是与设备厂商无关的通用规范。HAMi 当前仅支持通过 CDI 注入 NVIDIA GPU；本文中的 HAMi 配置和排错方法不适用于其他厂商设备。
+
+CDI 使用完全限定的设备名称标识设备：
+
+```text
+vendor.com/class=device-name
+```
+
+容器运行时收到设备名称后，会从 `/etc/cdi`、`/var/run/cdi` 等目录查找对应的 CDI spec，并将其中的配置写入容器的 OCI runtime spec。
+
+CDI 本身不负责设备调度、资源分配或配额管理。在 HAMi 的 NVIDIA CDI 集成中，调度器和 NVIDIA Device Plugin 仍然负责选择、分配 GPU；CDI 负责把已经分配的 GPU 及其运行环境注入容器。
+
+## CDI 解决什么问题
+
+复杂设备通常不能只靠挂载一个 `/dev` 设备节点完成注入。GPU 容器还可能需要驱动库、多个设备节点、环境变量和容器生命周期 hook。没有统一接口时，设备厂商需要分别适配不同的容器运行时，运行时中也容易出现厂商专用逻辑。
+
+CDI 将这些设备相关的 OCI 配置集中写入 spec，主要解决以下问题：
+
+- 使用统一格式描述设备节点、挂载、环境变量和 hook。
+- 使用稳定的设备名称在 Device Plugin 与容器运行时之间传递分配结果。
+- 减少设备注入逻辑对特定容器运行时实现的依赖。
+- 便于检查容器最终使用了哪些设备配置。
+
+## HAMi 如何为 NVIDIA GPU 使用 CDI
+
+为 NVIDIA Device Plugin 启用 `cdi-annotations` 后，HAMi 的设备注入流程如下：
+
+1. HAMi 调度器和 NVIDIA Device Plugin 完成 GPU 选择与分配。
+2. NVIDIA Device Plugin 生成 HAMi 使用的 CDI spec，并写入节点的 `/var/run/cdi` 目录。
+3. Device Plugin 通过 CDI annotation 返回分配到的设备名称。
+4. 支持 CDI 的容器运行时读取 annotation，查找对应的 CDI spec，并将设备配置加入容器的 OCI runtime spec。
+
+HAMi 生成的 NVIDIA GPU CDI kind 为：
+
+```text
+k8s.device-plugin.nvidia.com/gpu
+```
+
+## 前置条件
+
+启用 HAMi 的 NVIDIA CDI 支持前，需要确认以下条件：
+
+- 节点已经安装 NVIDIA 驱动和 NVIDIA Container Toolkit。
+- 容器运行时支持并已启用 CDI。
+- 容器运行时能够读取 `/var/run/cdi` 中的 CDI spec。
+- 已确认 NVIDIA 驱动和 `nvidia-ctk` 在节点上的实际路径。
+
+不同容器运行时的 CDI 配置方式，参阅 [CDI 上游配置说明](https://github.com/cncf-tags/container-device-interface#how-to-configure-cdi)。
+
+## 配置 Helm Chart
+
+将以下内容保存为 `values-cdi.yaml`：
+
+```yaml
+devicePlugin:
+  deviceListStrategy: "cdi-annotations"
+  nvidiaDriverRoot: "<driver-root>"
+  nvidiaHookPath: "<nvidia-ctk-path>"
+```
+
+- `devicePlugin.deviceListStrategy`：固定设置为 `cdi-annotations`，让 HAMi Device Plugin 通过 CDI annotation 传递设备名称。
+- `devicePlugin.nvidiaDriverRoot`：指定节点上的 NVIDIA 驱动根目录，HAMi 使用该目录发现驱动文件和设备节点。
+- `devicePlugin.nvidiaHookPath`：指定节点上实际执行的 `nvidia-ctk`，HAMi 将该路径写入生成的 CDI spec。
+
+根据部署方式替换两个路径占位符：
+
+- NVIDIA 驱动和 NVIDIA Container Toolkit 直接安装在宿主机：`<driver-root>` 使用 `/`，`<nvidia-ctk-path>` 使用 `/usr/bin/nvidia-ctk`。
+- GPU Operator 管理驱动和 NVIDIA Container Toolkit：`<driver-root>` 使用 `/run/nvidia/driver`，`<nvidia-ctk-path>` 使用 `/usr/local/nvidia/toolkit/nvidia-ctk`。
+
+:::warning
+
+两个路径参数必须按节点上的实际文件布局设置，不能直接套用另一种部署方式的值。`nvidiaHookPath` 必须是容器运行时所在节点能够执行的路径。
+
+:::
+
+安装或升级 HAMi：
+
+```bash
+helm upgrade --install hami hami-charts/hami \
+  --namespace kube-system \
+  --create-namespace \
+  --values values-cdi.yaml
+```
+
+## 验证配置
+
+先确认 Helm 使用了预期配置：
+
+```bash
+helm get values hami -n kube-system
+```
+
+然后在 GPU 节点上检查所选 `nvidia-ctk` 路径和 HAMi 生成的 CDI spec。下面以宿主机安装路径为例；GPU Operator 环境需要替换 `NVIDIA_CTK_PATH`：
+
+```bash
+NVIDIA_CTK_PATH=/usr/bin/nvidia-ctk
+sudo test -x "$NVIDIA_CTK_PATH"
+"$NVIDIA_CTK_PATH" --version
+sudo ls -l /var/run/cdi
+sudo jq . /var/run/cdi/k8s.device-plugin.nvidia.com-gpu.json
+```
+
+重点确认以下内容：
+
+- CDI kind 为 `k8s.device-plugin.nvidia.com/gpu`。
+- 分配使用的 GPU UUID 出现在设备列表中。
+- spec 中记录的设备节点和 hook 路径在节点上真实存在。
+
+最后检查 Device Plugin 日志：
+
+```bash
+kubectl logs -n kube-system \
+  -l app.kubernetes.io/component=hami-device-plugin \
+  --tail=200 | grep -i cdi
+```
+
+日志中不应出现 CDI spec 生成失败、设备无法解析或 hook 路径不存在等错误。
+
+## 常见问题
+
+### `failed to stat CDI host device "/dev/nvidia-modeset": no such file or directory`
+
+`devicePlugin.nvidiaDriverRoot` 与节点的驱动布局不一致。按前文的路径规则修正该参数，升级 HAMi，然后重新检查生成的 CDI spec。
+
+### `stderr: No help topic for 'disable-device-node-modification'`
+
+该错误表示实际执行的 NVIDIA Container Toolkit 不支持 CDI spec 中使用的 `disable-device-node-modification` 命令。该命令需要 NVIDIA Container Toolkit 1.18.0 或更高版本。
+
+将 NVIDIA Container Toolkit 升级到 1.18.0 或更高版本，并检查 CDI spec 中记录的 hook 路径及其实际版本。如果同时存在宿主机安装和 GPU Operator 容器化安装，需要分别检查；升级其中一套不会更新另一套。
+
+[GPU Operator v25.10.1](https://github.com/NVIDIA/gpu-operator/blob/v25.10.1/deployments/gpu-operator/values.yaml#L220-L224) 使用 NVIDIA Container Toolkit v1.18.1。相关问题参阅 [Toolkit version parser broken for casual kind demo user](https://github.com/kubernetes-sigs/dra-driver-nvidia-gpu/issues/568)。
+
+### `error running createContainer hook #0: fork/exec /usr/bin/nvidia-ctk: no such file or directory`
+
+CDI spec 中记录的 hook 路径在节点上不存在。确认 `nvidia-ctk` 的实际位置，修正 `devicePlugin.nvidiaHookPath`，升级 HAMi 以重新生成 CDI spec。
diff --git a/i18n/zh/docusaurus-plugin-content-docs/current/installation/how-to-use-hami-dra.md b/i18n/zh/docusaurus-plugin-content-docs/current/installation/how-to-use-hami-dra.md
index 7293b584..ba8593af 100644
--- a/i18n/zh/docusaurus-plugin-content-docs/current/installation/how-to-use-hami-dra.md
+++ b/i18n/zh/docusaurus-plugin-content-docs/current/installation/how-to-use-hami-dra.md
@@ -18,7 +18,7 @@ HAMi 已经提供了对 K8s [DRA](https://kubernetes.io/docs/concepts/scheduling
 ## 前提条件
 
 - Kubernetes 版本 >= 1.34 并且 DRA Consumable Capacity [featuregate](https://kubernetes.io/docs/reference/command-line-tools-reference/feature-gates/) 已启用
-- 底层容器运行时（例如 containerd 或 CRI-O）必须启用 [CDI](https://github.com/cncf-tags/container-device-interface?tab=readme-ov-file#how-to-configure-cdi)
+- 底层容器运行时（例如 containerd 或 CRI-O）必须启用 [CDI](./configure-cdi.md)
 - NVIDIA GPU 驱动版本 440 及以上
 
 ## 安装
diff --git a/i18n/zh/docusaurus-plugin-content-docs/current/installation/prerequisites.md b/i18n/zh/docusaurus-plugin-content-docs/current/installation/prerequisites.md
index 5bd5d3dd..4cf563d5 100644
--- a/i18n/zh/docusaurus-plugin-content-docs/current/installation/prerequisites.md
+++ b/i18n/zh/docusaurus-plugin-content-docs/current/installation/prerequisites.md
@@ -19,6 +19,8 @@ title: 前置条件
 
 参阅[安装 NVIDIA Container Toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html)。
 
+如需使用 CDI 注入 NVIDIA GPU，请在安装 HAMi 前参阅[为 HAMi 启用 NVIDIA CDI 支持](./configure-cdi.md)，确认容器运行时、驱动根目录和 NVIDIA Container Toolkit 路径。
+
 ### Debian 系统示例（Docker + containerd）
 
 #### 安装 `nvidia-container-toolkit`
diff --git a/i18n/zh/docusaurus-plugin-content-docs/version-v2.8.0/installation/configure-cdi.md b/i18n/zh/docusaurus-plugin-content-docs/version-v2.8.0/installation/configure-cdi.md
new file mode 100644
index 00000000..c1476b1a
--- /dev/null
+++ b/i18n/zh/docusaurus-plugin-content-docs/version-v2.8.0/installation/configure-cdi.md
@@ -0,0 +1,148 @@
+---
+title: 为 HAMi 启用 NVIDIA CDI 支持
+sidebar_label: NVIDIA CDI 支持
+translated: true
+---
+
+本文介绍 CDI 的作用，以及为 HAMi 配置 NVIDIA CDI 支持所需的 Helm Chart 参数。
+
+## CDI 是什么
+
+[CDI（Container Device Interface）](https://github.com/cncf-tags/container-device-interface)是一项容器设备接口规范。它使用 JSON 或 YAML 文件描述容器使用某个设备时需要应用的 OCI 配置，包括设备节点、文件挂载、环境变量和 OCI hook。
+
+CDI 是与设备厂商无关的通用规范。HAMi 当前仅支持通过 CDI 注入 NVIDIA GPU；本文中的 HAMi 配置和排错方法不适用于其他厂商设备。
+
+CDI 使用完全限定的设备名称标识设备：
+
+```text
+vendor.com/class=device-name
+```
+
+容器运行时收到设备名称后，会从 `/etc/cdi`、`/var/run/cdi` 等目录查找对应的 CDI spec，并将其中的配置写入容器的 OCI runtime spec。
+
+CDI 本身不负责设备调度、资源分配或配额管理。在 HAMi 的 NVIDIA CDI 集成中，调度器和 NVIDIA Device Plugin 仍然负责选择、分配 GPU；CDI 负责把已经分配的 GPU 及其运行环境注入容器。
+
+## CDI 解决什么问题
+
+复杂设备通常不能只靠挂载一个 `/dev` 设备节点完成注入。GPU 容器还可能需要驱动库、多个设备节点、环境变量和容器生命周期 hook。没有统一接口时，设备厂商需要分别适配不同的容器运行时，运行时中也容易出现厂商专用逻辑。
+
+CDI 将这些设备相关的 OCI 配置集中写入 spec，主要解决以下问题：
+
+- 使用统一格式描述设备节点、挂载、环境变量和 hook。
+- 使用稳定的设备名称在 Device Plugin 与容器运行时之间传递分配结果。
+- 减少设备注入逻辑对特定容器运行时实现的依赖。
+- 便于检查容器最终使用了哪些设备配置。
+
+## HAMi 如何为 NVIDIA GPU 使用 CDI
+
+为 NVIDIA Device Plugin 启用 `cdi-annotations` 后，HAMi 的设备注入流程如下：
+
+1. HAMi 调度器和 NVIDIA Device Plugin 完成 GPU 选择与分配。
+2. NVIDIA Device Plugin 生成 HAMi 使用的 CDI spec，并写入节点的 `/var/run/cdi` 目录。
+3. Device Plugin 通过 CDI annotation 返回分配到的设备名称。
+4. 支持 CDI 的容器运行时读取 annotation，查找对应的 CDI spec，并将设备配置加入容器的 OCI runtime spec。
+
+HAMi 生成的 NVIDIA GPU CDI kind 为：
+
+```text
+k8s.device-plugin.nvidia.com/gpu
+```
+
+## 前置条件
+
+启用 HAMi 的 NVIDIA CDI 支持前，需要确认以下条件：
+
+- 节点已经安装 NVIDIA 驱动和 NVIDIA Container Toolkit。
+- 容器运行时支持并已启用 CDI。
+- 容器运行时能够读取 `/var/run/cdi` 中的 CDI spec。
+- 已确认 NVIDIA 驱动和 `nvidia-ctk` 在节点上的实际路径。
+
+不同容器运行时的 CDI 配置方式，参阅 [CDI 上游配置说明](https://github.com/cncf-tags/container-device-interface#how-to-configure-cdi)。
+
+## 配置 Helm Chart
+
+将以下内容保存为 `values-cdi.yaml`：
+
+```yaml
+devicePlugin:
+  deviceListStrategy: "cdi-annotations"
+  nvidiaDriverRoot: "<driver-root>"
+  nvidiaHookPath: "<nvidia-ctk-path>"
+```
+
+- `devicePlugin.deviceListStrategy`：固定设置为 `cdi-annotations`，让 HAMi Device Plugin 通过 CDI annotation 传递设备名称。
+- `devicePlugin.nvidiaDriverRoot`：指定节点上的 NVIDIA 驱动根目录，HAMi 使用该目录发现驱动文件和设备节点。
+- `devicePlugin.nvidiaHookPath`：指定节点上实际执行的 `nvidia-ctk`，HAMi 将该路径写入生成的 CDI spec。
+
+根据部署方式替换两个路径占位符：
+
+- NVIDIA 驱动和 NVIDIA Container Toolkit 直接安装在宿主机：`<driver-root>` 使用 `/`，`<nvidia-ctk-path>` 使用 `/usr/bin/nvidia-ctk`。
+- GPU Operator 管理驱动和 NVIDIA Container Toolkit：`<driver-root>` 使用 `/run/nvidia/driver`，`<nvidia-ctk-path>` 使用 `/usr/local/nvidia/toolkit/nvidia-ctk`。
+
+:::warning
+
+两个路径参数必须按节点上的实际文件布局设置，不能直接套用另一种部署方式的值。`nvidiaHookPath` 必须是容器运行时所在节点能够执行的路径。
+
+:::
+
+安装或升级 HAMi：
+
+```bash
+helm upgrade --install hami hami-charts/hami \
+  --version 2.8.0 \
+  --namespace kube-system \
+  --create-namespace \
+  --values values-cdi.yaml
+```
+
+## 验证配置
+
+先确认 Helm 使用了预期配置：
+
+```bash
+helm get values hami -n kube-system
+```
+
+然后在 GPU 节点上检查所选 `nvidia-ctk` 路径和 HAMi 生成的 CDI spec。下面以宿主机安装路径为例；GPU Operator 环境需要替换 `NVIDIA_CTK_PATH`：
+
+```bash
+NVIDIA_CTK_PATH=/usr/bin/nvidia-ctk
+sudo test -x "$NVIDIA_CTK_PATH"
+"$NVIDIA_CTK_PATH" --version
+sudo ls -l /var/run/cdi
+sudo jq . /var/run/cdi/k8s.device-plugin.nvidia.com-gpu.json
+```
+
+重点确认以下内容：
+
+- CDI kind 为 `k8s.device-plugin.nvidia.com/gpu`。
+- 分配使用的 GPU UUID 出现在设备列表中。
+- spec 中记录的设备节点和 hook 路径在节点上真实存在。
+
+最后检查 Device Plugin 日志：
+
+```bash
+kubectl logs -n kube-system \
+  -l app.kubernetes.io/component=hami-device-plugin \
+  --tail=200 | grep -i cdi
+```
+
+日志中不应出现 CDI spec 生成失败、设备无法解析或 hook 路径不存在等错误。
+
+## 常见问题
+
+### `failed to stat CDI host device "/dev/nvidia-modeset": no such file or directory`
+
+`devicePlugin.nvidiaDriverRoot` 与节点的驱动布局不一致。按前文的路径规则修正该参数，升级 HAMi，然后重新检查生成的 CDI spec。
+
+### `stderr: No help topic for 'disable-device-node-modification'`
+
+该错误表示实际执行的 NVIDIA Container Toolkit 不支持 CDI spec 中使用的 `disable-device-node-modification` 命令。该命令需要 NVIDIA Container Toolkit 1.18.0 或更高版本。
+
+将 NVIDIA Container Toolkit 升级到 1.18.0 或更高版本，并检查 CDI spec 中记录的 hook 路径及其实际版本。如果同时存在宿主机安装和 GPU Operator 容器化安装，需要分别检查；升级其中一套不会更新另一套。
+
+[GPU Operator v25.10.1](https://github.com/NVIDIA/gpu-operator/blob/v25.10.1/deployments/gpu-operator/values.yaml#L220-L224) 使用 NVIDIA Container Toolkit v1.18.1。相关问题参阅 [Toolkit version parser broken for casual kind demo user](https://github.com/kubernetes-sigs/dra-driver-nvidia-gpu/issues/568)。
+
+### `error running createContainer hook #0: fork/exec /usr/bin/nvidia-ctk: no such file or directory`
+
+CDI spec 中记录的 hook 路径在节点上不存在。确认 `nvidia-ctk` 的实际位置，修正 `devicePlugin.nvidiaHookPath`，升级 HAMi 以重新生成 CDI spec。
diff --git a/i18n/zh/docusaurus-plugin-content-docs/version-v2.8.0/installation/how-to-use-hami-dra.md b/i18n/zh/docusaurus-plugin-content-docs/version-v2.8.0/installation/how-to-use-hami-dra.md
index a5fdcdfa..9bac135d 100644
--- a/i18n/zh/docusaurus-plugin-content-docs/version-v2.8.0/installation/how-to-use-hami-dra.md
+++ b/i18n/zh/docusaurus-plugin-content-docs/version-v2.8.0/installation/how-to-use-hami-dra.md
@@ -12,6 +12,7 @@ HAMi 已经提供了对 K8s [DRA](https://kubernetes.io/docs/concepts/scheduling
 ## 前提条件
 
 - Kubernetes 版本 >= 1.34 并且 DRA Consumable Capacity [featuregate](https://kubernetes.io/docs/reference/command-line-tools-reference/feature-gates/) 启用
+- 底层容器运行时（例如 containerd 或 CRI-O）必须启用 [CDI](./configure-cdi.md)
 
 ## 安装
 
diff --git a/i18n/zh/docusaurus-plugin-content-docs/version-v2.8.0/installation/prerequisites.md b/i18n/zh/docusaurus-plugin-content-docs/version-v2.8.0/installation/prerequisites.md
index 70966793..95b90de5 100644
--- a/i18n/zh/docusaurus-plugin-content-docs/version-v2.8.0/installation/prerequisites.md
+++ b/i18n/zh/docusaurus-plugin-content-docs/version-v2.8.0/installation/prerequisites.md
@@ -20,6 +20,8 @@ translated: true
 
 参阅：[https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html)
 
+如需使用 CDI 注入 NVIDIA GPU，请在安装 HAMi 前参阅[为 HAMi 启用 NVIDIA CDI 支持](./configure-cdi.md)，确认容器运行时、驱动根目录和 NVIDIA Container Toolkit 路径。
+
 ### 适用于基于 Debian 系统的 `Docker` 和 `containerd` 示例
 
 #### 安装 `nvidia-container-toolkit`
diff --git a/i18n/zh/docusaurus-plugin-content-docs/version-v2.9.0/installation/configure-cdi.md b/i18n/zh/docusaurus-plugin-content-docs/version-v2.9.0/installation/configure-cdi.md
new file mode 100644
index 00000000..717eb911
--- /dev/null
+++ b/i18n/zh/docusaurus-plugin-content-docs/version-v2.9.0/installation/configure-cdi.md
@@ -0,0 +1,148 @@
+---
+title: 为 HAMi 启用 NVIDIA CDI 支持
+sidebar_label: NVIDIA CDI 支持
+translated: true
+---
+
+本文介绍 CDI 的作用，以及为 HAMi 配置 NVIDIA CDI 支持所需的 Helm Chart 参数。
+
+## CDI 是什么
+
+[CDI（Container Device Interface）](https://github.com/cncf-tags/container-device-interface)是一项容器设备接口规范。它使用 JSON 或 YAML 文件描述容器使用某个设备时需要应用的 OCI 配置，包括设备节点、文件挂载、环境变量和 OCI hook。
+
+CDI 是与设备厂商无关的通用规范。HAMi 当前仅支持通过 CDI 注入 NVIDIA GPU；本文中的 HAMi 配置和排错方法不适用于其他厂商设备。
+
+CDI 使用完全限定的设备名称标识设备：
+
+```text
+vendor.com/class=device-name
+```
+
+容器运行时收到设备名称后，会从 `/etc/cdi`、`/var/run/cdi` 等目录查找对应的 CDI spec，并将其中的配置写入容器的 OCI runtime spec。
+
+CDI 本身不负责设备调度、资源分配或配额管理。在 HAMi 的 NVIDIA CDI 集成中，调度器和 NVIDIA Device Plugin 仍然负责选择、分配 GPU；CDI 负责把已经分配的 GPU 及其运行环境注入容器。
+
+## CDI 解决什么问题
+
+复杂设备通常不能只靠挂载一个 `/dev` 设备节点完成注入。GPU 容器还可能需要驱动库、多个设备节点、环境变量和容器生命周期 hook。没有统一接口时，设备厂商需要分别适配不同的容器运行时，运行时中也容易出现厂商专用逻辑。
+
+CDI 将这些设备相关的 OCI 配置集中写入 spec，主要解决以下问题：
+
+- 使用统一格式描述设备节点、挂载、环境变量和 hook。
+- 使用稳定的设备名称在 Device Plugin 与容器运行时之间传递分配结果。
+- 减少设备注入逻辑对特定容器运行时实现的依赖。
+- 便于检查容器最终使用了哪些设备配置。
+
+## HAMi 如何为 NVIDIA GPU 使用 CDI
+
+为 NVIDIA Device Plugin 启用 `cdi-annotations` 后，HAMi 的设备注入流程如下：
+
+1. HAMi 调度器和 NVIDIA Device Plugin 完成 GPU 选择与分配。
+2. NVIDIA Device Plugin 生成 HAMi 使用的 CDI spec，并写入节点的 `/var/run/cdi` 目录。
+3. Device Plugin 通过 CDI annotation 返回分配到的设备名称。
+4. 支持 CDI 的容器运行时读取 annotation，查找对应的 CDI spec，并将设备配置加入容器的 OCI runtime spec。
+
+HAMi 生成的 NVIDIA GPU CDI kind 为：
+
+```text
+k8s.device-plugin.nvidia.com/gpu
+```
+
+## 前置条件
+
+启用 HAMi 的 NVIDIA CDI 支持前，需要确认以下条件：
+
+- 节点已经安装 NVIDIA 驱动和 NVIDIA Container Toolkit。
+- 容器运行时支持并已启用 CDI。
+- 容器运行时能够读取 `/var/run/cdi` 中的 CDI spec。
+- 已确认 NVIDIA 驱动和 `nvidia-ctk` 在节点上的实际路径。
+
+不同容器运行时的 CDI 配置方式，参阅 [CDI 上游配置说明](https://github.com/cncf-tags/container-device-interface#how-to-configure-cdi)。
+
+## 配置 Helm Chart
+
+将以下内容保存为 `values-cdi.yaml`：
+
+```yaml
+devicePlugin:
+  deviceListStrategy: "cdi-annotations"
+  nvidiaDriverRoot: "<driver-root>"
+  nvidiaHookPath: "<nvidia-ctk-path>"
+```
+
+- `devicePlugin.deviceListStrategy`：固定设置为 `cdi-annotations`，让 HAMi Device Plugin 通过 CDI annotation 传递设备名称。
+- `devicePlugin.nvidiaDriverRoot`：指定节点上的 NVIDIA 驱动根目录，HAMi 使用该目录发现驱动文件和设备节点。
+- `devicePlugin.nvidiaHookPath`：指定节点上实际执行的 `nvidia-ctk`，HAMi 将该路径写入生成的 CDI spec。
+
+根据部署方式替换两个路径占位符：
+
+- NVIDIA 驱动和 NVIDIA Container Toolkit 直接安装在宿主机：`<driver-root>` 使用 `/`，`<nvidia-ctk-path>` 使用 `/usr/bin/nvidia-ctk`。
+- GPU Operator 管理驱动和 NVIDIA Container Toolkit：`<driver-root>` 使用 `/run/nvidia/driver`，`<nvidia-ctk-path>` 使用 `/usr/local/nvidia/toolkit/nvidia-ctk`。
+
+:::warning
+
+两个路径参数必须按节点上的实际文件布局设置，不能直接套用另一种部署方式的值。`nvidiaHookPath` 必须是容器运行时所在节点能够执行的路径。
+
+:::
+
+安装或升级 HAMi：
+
+```bash
+helm upgrade --install hami hami-charts/hami \
+  --version 2.9.0 \
+  --namespace kube-system \
+  --create-namespace \
+  --values values-cdi.yaml
+```
+
+## 验证配置
+
+先确认 Helm 使用了预期配置：
+
+```bash
+helm get values hami -n kube-system
+```
+
+然后在 GPU 节点上检查所选 `nvidia-ctk` 路径和 HAMi 生成的 CDI spec。下面以宿主机安装路径为例；GPU Operator 环境需要替换 `NVIDIA_CTK_PATH`：
+
+```bash
+NVIDIA_CTK_PATH=/usr/bin/nvidia-ctk
+sudo test -x "$NVIDIA_CTK_PATH"
+"$NVIDIA_CTK_PATH" --version
+sudo ls -l /var/run/cdi
+sudo jq . /var/run/cdi/k8s.device-plugin.nvidia.com-gpu.json
+```
+
+重点确认以下内容：
+
+- CDI kind 为 `k8s.device-plugin.nvidia.com/gpu`。
+- 分配使用的 GPU UUID 出现在设备列表中。
+- spec 中记录的设备节点和 hook 路径在节点上真实存在。
+
+最后检查 Device Plugin 日志：
+
+```bash
+kubectl logs -n kube-system \
+  -l app.kubernetes.io/component=hami-device-plugin \
+  --tail=200 | grep -i cdi
+```
+
+日志中不应出现 CDI spec 生成失败、设备无法解析或 hook 路径不存在等错误。
+
+## 常见问题
+
+### `failed to stat CDI host device "/dev/nvidia-modeset": no such file or directory`
+
+`devicePlugin.nvidiaDriverRoot` 与节点的驱动布局不一致。按前文的路径规则修正该参数，升级 HAMi，然后重新检查生成的 CDI spec。
+
+### `stderr: No help topic for 'disable-device-node-modification'`
+
+该错误表示实际执行的 NVIDIA Container Toolkit 不支持 CDI spec 中使用的 `disable-device-node-modification` 命令。该命令需要 NVIDIA Container Toolkit 1.18.0 或更高版本。
+
+将 NVIDIA Container Toolkit 升级到 1.18.0 或更高版本，并检查 CDI spec 中记录的 hook 路径及其实际版本。如果同时存在宿主机安装和 GPU Operator 容器化安装，需要分别检查；升级其中一套不会更新另一套。
+
+[GPU Operator v25.10.1](https://github.com/NVIDIA/gpu-operator/blob/v25.10.1/deployments/gpu-operator/values.yaml#L220-L224) 使用 NVIDIA Container Toolkit v1.18.1。相关问题参阅 [Toolkit version parser broken for casual kind demo user](https://github.com/kubernetes-sigs/dra-driver-nvidia-gpu/issues/568)。
+
+### `error running createContainer hook #0: fork/exec /usr/bin/nvidia-ctk: no such file or directory`
+
+CDI spec 中记录的 hook 路径在节点上不存在。确认 `nvidia-ctk` 的实际位置，修正 `devicePlugin.nvidiaHookPath`，升级 HAMi 以重新生成 CDI spec。
diff --git a/i18n/zh/docusaurus-plugin-content-docs/version-v2.9.0/installation/how-to-use-hami-dra.md b/i18n/zh/docusaurus-plugin-content-docs/version-v2.9.0/installation/how-to-use-hami-dra.md
index 7293b584..ba8593af 100644
--- a/i18n/zh/docusaurus-plugin-content-docs/version-v2.9.0/installation/how-to-use-hami-dra.md
+++ b/i18n/zh/docusaurus-plugin-content-docs/version-v2.9.0/installation/how-to-use-hami-dra.md
@@ -18,7 +18,7 @@ HAMi 已经提供了对 K8s [DRA](https://kubernetes.io/docs/concepts/scheduling
 ## 前提条件
 
 - Kubernetes 版本 >= 1.34 并且 DRA Consumable Capacity [featuregate](https://kubernetes.io/docs/reference/command-line-tools-reference/feature-gates/) 已启用
-- 底层容器运行时（例如 containerd 或 CRI-O）必须启用 [CDI](https://github.com/cncf-tags/container-device-interface?tab=readme-ov-file#how-to-configure-cdi)
+- 底层容器运行时（例如 containerd 或 CRI-O）必须启用 [CDI](./configure-cdi.md)
 - NVIDIA GPU 驱动版本 440 及以上
 
 ## 安装
diff --git a/i18n/zh/docusaurus-plugin-content-docs/version-v2.9.0/installation/prerequisites.md b/i18n/zh/docusaurus-plugin-content-docs/version-v2.9.0/installation/prerequisites.md
index a865a4a4..b90c5dd5 100644
--- a/i18n/zh/docusaurus-plugin-content-docs/version-v2.9.0/installation/prerequisites.md
+++ b/i18n/zh/docusaurus-plugin-content-docs/version-v2.9.0/installation/prerequisites.md
@@ -19,6 +19,8 @@ title: 前置条件
 
 参阅[安装 NVIDIA Container Toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html)。
 
+如需使用 CDI 注入 NVIDIA GPU，请在安装 HAMi 前参阅[为 HAMi 启用 NVIDIA CDI 支持](./configure-cdi.md)，确认容器运行时、驱动根目录和 NVIDIA Container Toolkit 路径。
+
 ### Debian 系统示例（Docker + containerd）
 
 #### 安装 `nvidia-container-toolkit`
diff --git a/sidebars.js b/sidebars.js
index db9a89b4..828ff4e7 100644
--- a/sidebars.js
+++ b/sidebars.js
@@ -49,6 +49,7 @@ module.exports = {
       },
       items: [
         "installation/prerequisites",
+        "installation/configure-cdi",
         "installation/online-installation",
         "installation/offline-installation",
         "installation/upgrade",
diff --git a/versioned_docs/version-v2.8.0/installation/configure-cdi.md b/versioned_docs/version-v2.8.0/installation/configure-cdi.md
new file mode 100644
index 00000000..d323f4d3
--- /dev/null
+++ b/versioned_docs/version-v2.8.0/installation/configure-cdi.md
@@ -0,0 +1,147 @@
+---
+title: Enable NVIDIA CDI support for HAMi
+sidebar_label: NVIDIA CDI support
+---
+
+This page explains what CDI does and which Helm chart values enable HAMi's NVIDIA CDI integration.
+
+## What is CDI?
+
+[CDI (Container Device Interface)](https://github.com/cncf-tags/container-device-interface) is a specification for describing the OCI configuration that a container needs to use a device. A CDI specification can define device nodes, mounts, environment variables, and OCI hooks in JSON or YAML.
+
+CDI is vendor-neutral. HAMi currently supports CDI device injection only for NVIDIA GPUs. The HAMi configuration and troubleshooting guidance on this page do not apply to devices from other vendors.
+
+CDI identifies each device by a fully qualified name:
+
+```text
+vendor.com/class=device-name
+```
+
+When a container runtime receives this name, it finds the matching CDI specification in a directory such as `/etc/cdi` or `/var/run/cdi` and applies the specified changes to the container's OCI runtime specification.
+
+CDI does not schedule devices, allocate resources, or enforce quotas. In HAMi's NVIDIA CDI integration, the scheduler and NVIDIA Device Plugin continue to select and allocate GPUs; CDI injects the allocated GPUs and their runtime requirements into containers.
+
+## What problems does CDI solve?
+
+Injecting a complex device often requires more than mounting a single `/dev` node. A GPU container may also need driver libraries, several device nodes, environment variables, and lifecycle hooks. Without a common interface, vendors must adapt this logic to individual container runtimes, and runtimes may need vendor-specific implementations.
+
+CDI moves these OCI changes into a common specification. It provides:
+
+- A standard format for device nodes, mounts, environment variables, and hooks.
+- Stable device names for passing allocation results from Device Plugins to container runtimes.
+- Less dependency on runtime-specific device injection implementations.
+- An inspectable record of the device configuration applied to a container.
+
+## How HAMi uses CDI for NVIDIA GPUs
+
+With `cdi-annotations` enabled for the NVIDIA Device Plugin, HAMi uses the following injection flow:
+
+1. The HAMi scheduler and NVIDIA Device Plugin select and allocate a GPU.
+2. The Device Plugin generates a CDI specification and writes it to `/var/run/cdi` on the node.
+3. The Device Plugin returns the allocated device name through a CDI annotation.
+4. A CDI-aware container runtime resolves the annotation and applies the matching CDI specification to the container's OCI runtime specification.
+
+HAMi uses the following CDI kind for NVIDIA GPUs:
+
+```text
+k8s.device-plugin.nvidia.com/gpu
+```
+
+## Prerequisites
+
+Before enabling NVIDIA CDI support in HAMi, confirm that:
+
+- The NVIDIA driver and NVIDIA Container Toolkit are installed on each GPU node.
+- The container runtime supports CDI and has CDI enabled.
+- The container runtime can read CDI specifications from `/var/run/cdi`.
+- The actual NVIDIA driver root and `nvidia-ctk` path on each node are known.
+
+For runtime-specific instructions, see the [upstream CDI configuration guide](https://github.com/cncf-tags/container-device-interface#how-to-configure-cdi).
+
+## Configure the Helm chart
+
+Save the following values as `values-cdi.yaml`:
+
+```yaml
+devicePlugin:
+  deviceListStrategy: "cdi-annotations"
+  nvidiaDriverRoot: "<driver-root>"
+  nvidiaHookPath: "<nvidia-ctk-path>"
+```
+
+- `devicePlugin.deviceListStrategy`: Set this to `cdi-annotations` so that the HAMi Device Plugin passes device names through CDI annotations.
+- `devicePlugin.nvidiaDriverRoot`: The NVIDIA driver root on the node. HAMi uses this directory to discover driver files and device nodes.
+- `devicePlugin.nvidiaHookPath`: The `nvidia-ctk` executable used on the node. HAMi writes this path into the generated CDI specification.
+
+Replace the two path placeholders according to the deployment:
+
+- Host-installed NVIDIA driver and Container Toolkit: use `/` for `<driver-root>` and `/usr/bin/nvidia-ctk` for `<nvidia-ctk-path>`.
+- GPU Operator-managed driver and Container Toolkit: use `/run/nvidia/driver` for `<driver-root>` and `/usr/local/nvidia/toolkit/nvidia-ctk` for `<nvidia-ctk-path>`.
+
+:::warning
+
+The two path values must match the actual node filesystem. Do not copy values from a different deployment model. `nvidiaHookPath` must point to an executable that is available to the container runtime on the node.
+
+:::
+
+Install or upgrade HAMi:
+
+```bash
+helm upgrade --install hami hami-charts/hami \
+  --version 2.8.0 \
+  --namespace kube-system \
+  --create-namespace \
+  --values values-cdi.yaml
+```
+
+## Verify the configuration
+
+First, confirm that Helm applied the expected values:
+
+```bash
+helm get values hami -n kube-system
+```
+
+On a GPU node, check the selected `nvidia-ctk` executable and the CDI specification generated by HAMi. The following example uses the host-installed path; replace `NVIDIA_CTK_PATH` in GPU Operator environments:
+
+```bash
+NVIDIA_CTK_PATH=/usr/bin/nvidia-ctk
+sudo test -x "$NVIDIA_CTK_PATH"
+"$NVIDIA_CTK_PATH" --version
+sudo ls -l /var/run/cdi
+sudo jq . /var/run/cdi/k8s.device-plugin.nvidia.com-gpu.json
+```
+
+Confirm that:
+
+- The CDI kind is `k8s.device-plugin.nvidia.com/gpu`.
+- The allocated GPU UUID is present in the device list.
+- Every device node and hook path in the specification exists on the node.
+
+Finally, check the Device Plugin logs:
+
+```bash
+kubectl logs -n kube-system \
+  -l app.kubernetes.io/component=hami-device-plugin \
+  --tail=200 | grep -i cdi
+```
+
+The logs should not contain errors about CDI specification generation, unresolved devices, or missing hook paths.
+
+## Troubleshooting
+
+### `failed to stat CDI host device "/dev/nvidia-modeset": no such file or directory`
+
+`devicePlugin.nvidiaDriverRoot` does not match the driver layout on the node. Correct the value using the path rules above, upgrade HAMi, and inspect the regenerated CDI specification.
+
+### `stderr: No help topic for 'disable-device-node-modification'`
+
+The NVIDIA Container Toolkit executable referenced by the CDI specification does not support the `disable-device-node-modification` command. This command requires NVIDIA Container Toolkit 1.18.0 or later.
+
+Upgrade NVIDIA Container Toolkit and verify both the hook path recorded in the CDI specification and the version of the executable at that path. Host-installed and GPU Operator-managed Toolkit installations are independent; upgrading one does not update the other.
+
+[GPU Operator v25.10.1](https://github.com/NVIDIA/gpu-operator/blob/v25.10.1/deployments/gpu-operator/values.yaml#L220-L224) uses NVIDIA Container Toolkit v1.18.1. For related details, see [Toolkit version parser broken for casual kind demo user](https://github.com/kubernetes-sigs/dra-driver-nvidia-gpu/issues/568).
+
+### `error running createContainer hook #0: fork/exec /usr/bin/nvidia-ctk: no such file or directory`
+
+The hook path recorded in the CDI specification does not exist on the node. Locate the actual `nvidia-ctk` executable, correct `devicePlugin.nvidiaHookPath`, and upgrade HAMi to regenerate the CDI specification.
diff --git a/versioned_docs/version-v2.8.0/installation/how-to-use-hami-dra.md b/versioned_docs/version-v2.8.0/installation/how-to-use-hami-dra.md
index 8eb39838..70661d72 100644
--- a/versioned_docs/version-v2.8.0/installation/how-to-use-hami-dra.md
+++ b/versioned_docs/version-v2.8.0/installation/how-to-use-hami-dra.md
@@ -12,6 +12,7 @@ By installing the [HAMi DRA webhook](https://github.com/Project-HAMi/HAMi-DRA) i
 ## Prerequisites
 
 * Kubernetes version >= 1.34 with DRA Consumable Capacity [featuregate](https://kubernetes.io/docs/reference/command-line-tools-reference/feature-gates/) enabled
+* [CDI](./configure-cdi.md) must be enabled in the underlying container runtime (such as containerd or CRI-O)
 
 ## Installation
 
diff --git a/versioned_docs/version-v2.8.0/installation/prerequisites.md b/versioned_docs/version-v2.8.0/installation/prerequisites.md
index 01519388..596c9491 100644
--- a/versioned_docs/version-v2.8.0/installation/prerequisites.md
+++ b/versioned_docs/version-v2.8.0/installation/prerequisites.md
@@ -19,6 +19,8 @@ This README assumes pre-installation of NVIDIA drivers and the `nvidia-container
 
 For details see [Installing the NVIDIA Container Toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html).
 
+To inject NVIDIA GPUs through CDI, follow [Enable NVIDIA CDI support for HAMi](./configure-cdi.md) before installing HAMi. Verify the container runtime, driver root, and NVIDIA Container Toolkit path first.
+
 ### Example for debian-based systems with Docker and containerd
 
 #### Install the `nvidia-container-toolkit`
diff --git a/versioned_docs/version-v2.9.0/installation/configure-cdi.md b/versioned_docs/version-v2.9.0/installation/configure-cdi.md
new file mode 100644
index 00000000..772b7667
--- /dev/null
+++ b/versioned_docs/version-v2.9.0/installation/configure-cdi.md
@@ -0,0 +1,147 @@
+---
+title: Enable NVIDIA CDI support for HAMi
+sidebar_label: NVIDIA CDI support
+---
+
+This page explains what CDI does and which Helm chart values enable HAMi's NVIDIA CDI integration.
+
+## What is CDI?
+
+[CDI (Container Device Interface)](https://github.com/cncf-tags/container-device-interface) is a specification for describing the OCI configuration that a container needs to use a device. A CDI specification can define device nodes, mounts, environment variables, and OCI hooks in JSON or YAML.
+
+CDI is vendor-neutral. HAMi currently supports CDI device injection only for NVIDIA GPUs. The HAMi configuration and troubleshooting guidance on this page do not apply to devices from other vendors.
+
+CDI identifies each device by a fully qualified name:
+
+```text
+vendor.com/class=device-name
+```
+
+When a container runtime receives this name, it finds the matching CDI specification in a directory such as `/etc/cdi` or `/var/run/cdi` and applies the specified changes to the container's OCI runtime specification.
+
+CDI does not schedule devices, allocate resources, or enforce quotas. In HAMi's NVIDIA CDI integration, the scheduler and NVIDIA Device Plugin continue to select and allocate GPUs; CDI injects the allocated GPUs and their runtime requirements into containers.
+
+## What problems does CDI solve?
+
+Injecting a complex device often requires more than mounting a single `/dev` node. A GPU container may also need driver libraries, several device nodes, environment variables, and lifecycle hooks. Without a common interface, vendors must adapt this logic to individual container runtimes, and runtimes may need vendor-specific implementations.
+
+CDI moves these OCI changes into a common specification. It provides:
+
+- A standard format for device nodes, mounts, environment variables, and hooks.
+- Stable device names for passing allocation results from Device Plugins to container runtimes.
+- Less dependency on runtime-specific device injection implementations.
+- An inspectable record of the device configuration applied to a container.
+
+## How HAMi uses CDI for NVIDIA GPUs
+
+With `cdi-annotations` enabled for the NVIDIA Device Plugin, HAMi uses the following injection flow:
+
+1. The HAMi scheduler and NVIDIA Device Plugin select and allocate a GPU.
+2. The Device Plugin generates a CDI specification and writes it to `/var/run/cdi` on the node.
+3. The Device Plugin returns the allocated device name through a CDI annotation.
+4. A CDI-aware container runtime resolves the annotation and applies the matching CDI specification to the container's OCI runtime specification.
+
+HAMi uses the following CDI kind for NVIDIA GPUs:
+
+```text
+k8s.device-plugin.nvidia.com/gpu
+```
+
+## Prerequisites
+
+Before enabling NVIDIA CDI support in HAMi, confirm that:
+
+- The NVIDIA driver and NVIDIA Container Toolkit are installed on each GPU node.
+- The container runtime supports CDI and has CDI enabled.
+- The container runtime can read CDI specifications from `/var/run/cdi`.
+- The actual NVIDIA driver root and `nvidia-ctk` path on each node are known.
+
+For runtime-specific instructions, see the [upstream CDI configuration guide](https://github.com/cncf-tags/container-device-interface#how-to-configure-cdi).
+
+## Configure the Helm chart
+
+Save the following values as `values-cdi.yaml`:
+
+```yaml
+devicePlugin:
+  deviceListStrategy: "cdi-annotations"
+  nvidiaDriverRoot: "<driver-root>"
+  nvidiaHookPath: "<nvidia-ctk-path>"
+```
+
+- `devicePlugin.deviceListStrategy`: Set this to `cdi-annotations` so that the HAMi Device Plugin passes device names through CDI annotations.
+- `devicePlugin.nvidiaDriverRoot`: The NVIDIA driver root on the node. HAMi uses this directory to discover driver files and device nodes.
+- `devicePlugin.nvidiaHookPath`: The `nvidia-ctk` executable used on the node. HAMi writes this path into the generated CDI specification.
+
+Replace the two path placeholders according to the deployment:
+
+- Host-installed NVIDIA driver and Container Toolkit: use `/` for `<driver-root>` and `/usr/bin/nvidia-ctk` for `<nvidia-ctk-path>`.
+- GPU Operator-managed driver and Container Toolkit: use `/run/nvidia/driver` for `<driver-root>` and `/usr/local/nvidia/toolkit/nvidia-ctk` for `<nvidia-ctk-path>`.
+
+:::warning
+
+The two path values must match the actual node filesystem. Do not copy values from a different deployment model. `nvidiaHookPath` must point to an executable that is available to the container runtime on the node.
+
+:::
+
+Install or upgrade HAMi:
+
+```bash
+helm upgrade --install hami hami-charts/hami \
+  --version 2.9.0 \
+  --namespace kube-system \
+  --create-namespace \
+  --values values-cdi.yaml
+```
+
+## Verify the configuration
+
+First, confirm that Helm applied the expected values:
+
+```bash
+helm get values hami -n kube-system
+```
+
+On a GPU node, check the selected `nvidia-ctk` executable and the CDI specification generated by HAMi. The following example uses the host-installed path; replace `NVIDIA_CTK_PATH` in GPU Operator environments:
+
+```bash
+NVIDIA_CTK_PATH=/usr/bin/nvidia-ctk
+sudo test -x "$NVIDIA_CTK_PATH"
+"$NVIDIA_CTK_PATH" --version
+sudo ls -l /var/run/cdi
+sudo jq . /var/run/cdi/k8s.device-plugin.nvidia.com-gpu.json
+```
+
+Confirm that:
+
+- The CDI kind is `k8s.device-plugin.nvidia.com/gpu`.
+- The allocated GPU UUID is present in the device list.
+- Every device node and hook path in the specification exists on the node.
+
+Finally, check the Device Plugin logs:
+
+```bash
+kubectl logs -n kube-system \
+  -l app.kubernetes.io/component=hami-device-plugin \
+  --tail=200 | grep -i cdi
+```
+
+The logs should not contain errors about CDI specification generation, unresolved devices, or missing hook paths.
+
+## Troubleshooting
+
+### `failed to stat CDI host device "/dev/nvidia-modeset": no such file or directory`
+
+`devicePlugin.nvidiaDriverRoot` does not match the driver layout on the node. Correct the value using the path rules above, upgrade HAMi, and inspect the regenerated CDI specification.
+
+### `stderr: No help topic for 'disable-device-node-modification'`
+
+The NVIDIA Container Toolkit executable referenced by the CDI specification does not support the `disable-device-node-modification` command. This command requires NVIDIA Container Toolkit 1.18.0 or later.
+
+Upgrade NVIDIA Container Toolkit and verify both the hook path recorded in the CDI specification and the version of the executable at that path. Host-installed and GPU Operator-managed Toolkit installations are independent; upgrading one does not update the other.
+
+[GPU Operator v25.10.1](https://github.com/NVIDIA/gpu-operator/blob/v25.10.1/deployments/gpu-operator/values.yaml#L220-L224) uses NVIDIA Container Toolkit v1.18.1. For related details, see [Toolkit version parser broken for casual kind demo user](https://github.com/kubernetes-sigs/dra-driver-nvidia-gpu/issues/568).
+
+### `error running createContainer hook #0: fork/exec /usr/bin/nvidia-ctk: no such file or directory`
+
+The hook path recorded in the CDI specification does not exist on the node. Locate the actual `nvidia-ctk` executable, correct `devicePlugin.nvidiaHookPath`, and upgrade HAMi to regenerate the CDI specification.
diff --git a/versioned_docs/version-v2.9.0/installation/how-to-use-hami-dra.md b/versioned_docs/version-v2.9.0/installation/how-to-use-hami-dra.md
index 2695115e..6db129c5 100644
--- a/versioned_docs/version-v2.9.0/installation/how-to-use-hami-dra.md
+++ b/versioned_docs/version-v2.9.0/installation/how-to-use-hami-dra.md
@@ -18,7 +18,7 @@ HAMi has provided support for K8s [DRA](https://kubernetes.io/docs/concepts/sche
 ## Prerequisites
 
 - Kubernetes version >= 1.34 with DRA Consumable Capacity [featuregate](https://kubernetes.io/docs/reference/command-line-tools-reference/feature-gates/) enabled
-- [CDI](https://github.com/cncf-tags/container-device-interface?tab=readme-ov-file#how-to-configure-cdi) must be enabled in the underlying container runtime (such as containerd or CRI-O)
+- [CDI](./configure-cdi.md) must be enabled in the underlying container runtime (such as containerd or CRI-O)
 - NVIDIA GPU Driver 440 or later
 
 ## Installation
diff --git a/versioned_docs/version-v2.9.0/installation/prerequisites.md b/versioned_docs/version-v2.9.0/installation/prerequisites.md
index 24966331..cc43b3e6 100644
--- a/versioned_docs/version-v2.9.0/installation/prerequisites.md
+++ b/versioned_docs/version-v2.9.0/installation/prerequisites.md
@@ -19,6 +19,8 @@ This guide assumes pre-installation of NVIDIA drivers and the `nvidia-container-
 
 For details see [Installing the NVIDIA Container Toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html).
 
+To inject NVIDIA GPUs through CDI, follow [Enable NVIDIA CDI support for HAMi](./configure-cdi.md) before installing HAMi. Verify the container runtime, driver root, and NVIDIA Container Toolkit path first.
+
 ### Example for debian-based systems with Docker and containerd
 
 #### Install the `nvidia-container-toolkit`
diff --git a/versioned_sidebars/version-v2.8.0-sidebars.json b/versioned_sidebars/version-v2.8.0-sidebars.json
index f6860d80..fdae2454 100644
--- a/versioned_sidebars/version-v2.8.0-sidebars.json
+++ b/versioned_sidebars/version-v2.8.0-sidebars.json
@@ -31,6 +31,7 @@
       "label": "Installation",
       "items": [
         "installation/prerequisites",
+        "installation/configure-cdi",
         "installation/online-installation",
         "installation/offline-installation",
         "installation/upgrade",
diff --git a/versioned_sidebars/version-v2.9.0-sidebars.json b/versioned_sidebars/version-v2.9.0-sidebars.json
index 43434f82..6d0cd867 100644
--- a/versioned_sidebars/version-v2.9.0-sidebars.json
+++ b/versioned_sidebars/version-v2.9.0-sidebars.json
@@ -54,6 +54,7 @@
       },
       "items": [
         "installation/prerequisites",
+        "installation/configure-cdi",
         "installation/online-installation",
         "installation/offline-installation",
         "installation/upgrade",